Won't Discuss
Topics the AI refuses to engage on — legal advice, medical advice, competitor comparisons, anything that should be handed to a human. The AI acknowledges and redirects; it never improvises an answer.
Won't Discuss is the list of topics where your AI should stop, acknowledge it can't help with this specifically, and either redirect or capture the contact for human follow-up. It's the safety net for everything you absolutely don't want your AI improvising on.
Where to find it
Won't Discuss has its own page at /dashboard/knowledge/wont-discuss. You reach it from the "Won't Discuss" chip in the header of both the Knowledge Base page and the Receptionist page — they're mirrored, so either works.
One list, reached from two places
Won't Discuss has a single dedicated page. The "Won't discuss" chip on the Knowledge Base header and the matching chip on the Receptionist page both lead to it, and both edit the same list — there's no separate copy to keep in sync.
What belongs in Won't Discuss
- Anything regulated — specific legal advice, medical advice, financial guarantees, tax positions.
- Anything where being wrong has real consequences — quoted prices on custom work, definitive turnaround commitments, warranty interpretations.
- Anything outside your business — competitor comparisons, third-party recommendations, unrelated industry chatter.
- Anything personal to the caller you don't want logged — payment card numbers, social security numbers, medical history (your AI will still flag these in transcripts if a caller volunteers them, but Won't Discuss prevents the AI from prompting for them).
How a Won't Discuss entry behaves on a call
When a caller raises a topic on the list, your AI does three things in sequence: acknowledges the question without restating the sensitive part, briefly explains it can't help with this specifically, and offers a path forward (callback, email, transfer to a person). It does not improvise an answer. It does not guess. The redirect is handled by the AI from the situation — you don't configure a separate response for each topic.
Example
"I want to know if I can sue my landlord." → "That's a legal question I'm not able to advise on, but I can have an attorney from our office call you back. What's the best number to reach you on?"
How to use it
- 1Add topics in plain English. Each entry is just a topic name — "Specific legal advice", "competitor pricing", "medical diagnosis". Type one and press Enter, or click a suggested chip.
- 2Use the "Suggest for my business" button if you want a starting set. Your AI proposes blocked topics tailored to your industry; review them and remove any you don't want.
- 3Test by calling and raising each topic yourself. Confirm the AI redirects gracefully — neither cold-shouldering the caller nor leaking into actually answering.
- 4Review after the first 50 calls. Owners often discover topics they didn't think to list but the AI handled fine — keep those off the list. Add the ones that surprised you in the wrong direction.
Won't Discuss is broader than Rules — use it deliberately
A rule says "handle this topic this way." Won't Discuss says "stop and hand off." The list should be short. Over-using Won't Discuss makes the AI feel evasive on calls it could have handled with a Rule + Knowledge entry instead.
Common questions
My AI gave legal advice anyway.
Open the transcript. Either the topic wasn't specific enough (the caller used a phrasing the topic name didn't cover), or the AI answered before recognizing the topic. Add a more specific entry covering the actual phrasing.
Can I add a topic for "anything I haven't added to the KB"?
No — Won't Discuss is for topics you don't want answered. For everything else, the AI defaults to "I don't have that information, can I have someone follow up?" because of the never-guess contract. The right move for unknown topics is to add KB entries, not Won't Discuss entries.
What's the difference between Won't Discuss and a Rule that says "always say no to X"?
Won't Discuss is intent-level — the AI recognizes the topic and refuses. A Rule is response-level — the AI engages but with constraints. If the topic should never be engaged on at all, Won't Discuss. If it should be engaged but bounded, a Rule.
Can I control exactly what the AI says when it redirects?
Not per topic — each Won't Discuss entry is just the topic name, and the AI generates the acknowledgment and redirect itself, drawing on your Knowledge Base (a callback contact, the right person to reach). If a specific topic needs a specific scripted handling, write a Rule for it instead.
Knowledge Base
For topics you want your AI to answer — the opposite lever from Won't Discuss.
Rules
For topics that should be answered but shaped a specific way.
Related articles
Knowledge Base
Everything your AI knows about your business — services, hours, pricing, policies. A state model the AI consults, not a doc store. The more you fill in, the more your AI answers confidently.
Rules
Instructions for how the AI should handle situations — not what it knows, but how it acts. Use Rules when the right response depends on context, not on a specific fact.
Still stuck? Email support — we usually respond within a business day.